Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Long-context evaluation on RULER 32k (Zero-shot transfer)
Loading...
37.2
Accuracy
Still
18.064
23.032
28
32.968
Jun 5, 2026
Accuracy
Updated 1mo ago
Evaluation Results
Method
Method
Links
Accuracy
Still
K (Cache budget)=2048,...
2026.06
37.2
Still
K (Cache budget)=1024,...
2026.06
34
Still
K (Cache budget)=328,...
2026.06
33.7
Still
K (Cache budget)=164,...
2026.06
33.3
Still
K (Cache budget)=512,...
2026.06
29.6
KV-Distill
K (Cache budget)=2048,...
2026.06
24.1
KV-Distill
K (Cache budget)=1024,...
2026.06
22.7
KV-Distill
K (Cache budget)=164,...
2026.06
22.7
Still
K (Cache budget)=82, N...
2026.06
22.1
KV-Distill
K (Cache budget)=328,...
2026.06
22
KV-Distill
K (Cache budget)=512,...
2026.06
21.3
KV-Distill
K (Cache budget)=82, N...
2026.06
18.8
Feedback
Search any
task
Search any
task